454 Serviss — Internal Temperature Gradient of Metals. 



side of the thermopile, and the copper was bent twice to pass 

 around the ring-. The edges of the mica were marked at 

 regular intervals and very slightly notched to hold the wires 

 in position. The first and the last members of the thermopile 

 were, of course, copper, and were soldered directly to longer 

 pieces of the same wire. These were slipped through small 

 rubber tubing, coiled twice around the inside of the ther- 

 mostat at the level of the thermopile, and then through 

 85 cm of copper tubing passing through the ice, and then 

 soldered to the galvanometer leads. Thus fully a meter and 

 a half of each lead wire was at zero, adequately protecting the 

 end- junctions of the thermopile from heating by conduction 

 through them. In Dr. Thwing's apparatus, these wires came 

 in through a short tube between the thermostat and the ice 

 vat. Now it is probable that any constant error arising from 

 this source would tend to warm the outer junction rather than 

 the inner one, which is just the reverse of what Dr. Thwing 

 found. Therefore, if in spite of the possible masking of the 

 effect by conduction along his short lead wires, Dr. Thwing is 

 able to find an excess of temperature at the inner row, then 

 that excess ought to be as great, if not greater, when, as here, 

 this source of error is certainly eliminated. 



After the thermal couples were all in place on the mica 

 frame, they were painted with a thin coat of asphaltum var- 

 nish and baked in an electric oven for several hours at 120° C. 

 The insulation was found to be defective in a few places, but 

 after a second treatment stood a severe test. This varnish 

 proved a very satisfactory insulator when used for two years 

 on the main bars of the Thomson effect researches of Pro- 

 fessor E. H. Hall, in which the writer had a part.* It was 

 accordingly adopted in preference to the beeswax and rosin 

 used by Dr. Thwing, which may have contained acid, since 

 there is no evidence to show that it was especially prepared. 

 Furthermore, the use of asphaltum and thinner wires enabled 

 me to make a thermopile only a third as thick as Dr. Thwing's, 

 which is of course an advantage. And finally there was a 

 mechanical advantage. Although the couples with large 

 joints had been rejected, still the junctions were unavoid- 

 ably slight^ larger than the wires themselves, and the weight 

 of the upper cylinder rested upon these two hundred joints. 

 This insulation was a thin, smooth, hard coating which was 

 not rubbed off or squeezed aside by this pressure. 



There remains some possibility of disturbing effects due to 

 the circulation of air in the channels between the wires. To 

 prevent this, three rings of very thin mica (a, h, c in fig. 1) 



*Proc. Amer. Acad., xli, 23-55, 1905, and xlii, 595-626, 1907. 



